SPR 22-630, Pilot Study on Improving Crash Data Accuracy in Kentucky Through University Collaboration

PROBLEM STATEMENT Crash data are essential for conducting safety investigations and recommending changes in roadway geometry. But crash data sometimes suffer from coding errors — especially those related to crash types — which could prompt the decision to make erroneous safety improvements. Other crash data errors include incorrect vehicle directions, injury codes, and crash locations. For this project, three universities (University of Kentucky, University of Louisville, and Western Kentucky University) will collaborate to improve the accuracy of Kentucky crash data and correct miscoded crash data (e.g., crash types, vehicle maneuvers, injury codes). Text mining and machine learning will be used to identify specific keywords in crash report narratives and compare findings with stored crash data. OBJECTIVES • Improve the quality and accuracy of various elements in Kentucky’s crash data • Strengthen collaboration between participating universities by integrating research resources, which will help the institutions better serve Kentucky’s traffic safety community.